- Microsoft’s AI diagnostic system, called MAI-DxO, correctly diagnosed complex medical cases 83% of the time in a simulation using cases from the New England Journal of Medicine, versus 36% for primary care doctors and general internists working alone, according to a study published in Science and reported by TechCrunch.
- Microsoft AI CEO Mustafa Suleyman called the result a step toward “medical superintelligence,” though the comparison excluded specialists and physician teams who normally handle cases this complex.
- MAI-DxO coordinates several AI agents that act like a panel of doctors, questioning a patient, ordering tests, and debating a diagnosis; its best results came when paired with OpenAI’s GPT-5 and GPT-5.5 models.
- The study’s own data suggests much of the accuracy gain comes from the underlying AI models themselves, with Microsoft’s orchestration layer adding smaller additional gains on top.
What Happened
Microsoft published a study in Science showing its MAI-DxO diagnostic system correctly solved complex medical cases 83% of the time in simulation, more than double the 36% rate for primary care doctors and general internists working individually with reference materials, TechCrunch reported on October 22, 2026.
Why It Matters
Dominic King, Microsoft AI’s vice president of health, said “there is something very substantial and exciting here,” framing the result as a response to a widening gap between demand for medical expertise and its supply. Microsoft now joins Google and OpenAI in publishing claims that AI systems can match or exceed doctors on diagnostic accuracy, following Microsoft’s earlier claim that MAI-DxO got 85.5% of diagnoses right.
Technical Details
MAI-DxO was tested across models from OpenAI, Google, Anthropic, Meta, and DeepSeek. Results varied by case type: GPT-5.5 paired with DxO improved from 72% to nearly 87% on complex cases at a slightly lower cost, but showed little gain on emergency cases. The orchestration helped weaker models most; Gemini 2.5 Flash scored 46.4% alone but nearly 69% with DxO.
Who’s Affected
Healthcare systems facing clinician shortages are the intended audience. Suleyman said Microsoft is training a dedicated healthcare foundation model with the Mayo Clinic, for release next year, building on the partnership the two previously announced to build a frontier AI model for healthcare.
What’s Next
Suleyman said he expects “enormous leaps” by 2027 as the Mayo Clinic model matures, though the study notes gains shrink as models near their ceiling, meaning future progress may come more from new base models than orchestration.
